Full read-write Claude connector for Databricks
Let Claude run governed reads and writes against your Databricks lakehouse
Stop ending every analysis at a SELECT and hand-running the follow-up yourself. Claude queries, inspects schemas, and applies the change where the data already lives: governed inserts, table edits, and job runs, all behind your approval.
Databricks is a lakehouse platform where teams run SQL, manage tables, and operate jobs and notebooks on top of their data. People want Claude to do more than read query results: to run governed writes against the warehouse and act on the platform itself.
What is a Databricks MCP server?
A Databricks MCP server is a small hosted service that exposes Databricks's data and actions to Claude over the Model Context Protocol, so Claude can records, tables, and queries inside your own account. We build it as a custom connector with separate read and write tools, and destructive actions stay gated behind your approval.
What can Databricks's current connector do today?
Today, Databricks exposes MCP and SQL endpoints, but a general Claude connection tends to be read-leaning: Claude can run queries and inspect schemas, while writes like inserts, table changes, and job runs are limited or absent. A custom Databricks connector adds those write actions with scoped, human-in-the-loop controls.
What does full read-write Databricks access unlock?
Full read-write Databricks access lets Claude run governed inserts and updates, create or alter tables in Unity Catalog, trigger jobs and pipelines, and manage warehouses, all behind your approval. The analysis that used to end at a SELECT can now apply the change where the data already lives.
- Run a parameterized INSERT or MERGE into a target Delta table after you approve it
- Create or alter a table or schema in Unity Catalog
- Trigger a job run or workflow and report back when it finishes
- Start, stop, or resize a SQL warehouse to match the workload
- Grant or revoke a scoped permission on a catalog object
- Update or tag table metadata to keep the catalog accurate
Can Claude write to Databricks tables, not just query them?
Yes. A custom Databricks connector exposes scoped write tools, so Claude can run a parameterized INSERT or MERGE into a target Delta table, or create and alter a table or schema in Unity Catalog. Reads and writes are separate tools, and the connector only ever has the grants your account does.
Can Claude trigger a job or pipeline run in Databricks?
Claude can trigger a job run or workflow and report back when it finishes, and start, stop, or resize a SQL warehouse to match the workload. Job triggers are gated write tools, so Claude prepares the run and it executes only after you approve, keeping outbound actions under your control.
Can the connector access more than my Databricks role allows?
No. The connector authenticates as your OAuth identity or a scoped service principal and inherits Unity Catalog permissions. It can never touch a catalog, schema, or table your own role cannot, and access can be narrowed further per tool and locked to your workspace with IP allowlisting.
How a Databricks connector works
We build a small MCP server that authenticates to Databricks with OAuth or a scoped service principal and the narrowest Unity Catalog grants the work needs. Read tools (query, inspect schema) and write tools (insert, alter, run job) are separated, and destructive or outbound actions like drops and job triggers require your confirmation. The connector never has more access than the account it runs as, and it can be locked to your workspace with IP allowlisting.
Frequently asked questions
Can Claude write to Databricks tables, not just run SELECTs?
Will Claude drop a table or run a job without my approval?
Can the connector access more than my Databricks role allows?
Related connectors
Tell us what you run. We'll build the connector.
Pick your tool, choose what Claude should do, and get an instant scope and price. No call required.